blk-settings: round down io_opt to at least 4K

Commit Message

Mikulas Patocka Jan. 20, 2025, 3:16 p.m. UTC

Some SATA SSDs and most NVMe SSDs report physical block size 512 bytes,
but they use 4K remapping table internally and they do slow
read-modify-write cycle for requests that are not aligned on 4K boundary.
Therefore, io_opt should be aligned on 4K.

Signed-off-by: Mikulas Patocka <[e-mail]>
Fixes: a23634644afc ("block: take io_opt and io_min into account for max_sectors")
Fixes: 9c0ba14828d6 ("blk-settings: round down io_opt to physical_block_size")
Cc: [e-mail]	# v6.11+

---
 block/blk-settings.c |    6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

Patch

Index: linux-2.6/block/blk-settings.c
===================================================================
--- linux-2.6.orig/block/blk-settings.c	2025-01-03 21:10:56.000000000 +0100
+++ linux-2.6/block/blk-settings.c	2025-01-20 15:59:13.000000000 +0100
@@ -269,8 +269,12 @@  int blk_validate_limits(struct queue_lim
 	 * The optimal I/O size may not be aligned to physical block size
 	 * (because it may be limited by dma engines which have no clue about
 	 * block size of the disks attached to them), so we round it down here.
+	 *
+	 * Note that some SSDs erroneously report physical_block_size 512
+	 * despite the fact that they have remapping table granularity 4K and
+	 * they perform read-modify-write for unaligned requests.
 	 */
-	lim->io_opt = round_down(lim->io_opt, lim->physical_block_size);
+	lim->io_opt = round_down(lim->io_opt, max(4096, lim->physical_block_size));
 
 	/*
 	 * max_hw_sectors has a somewhat weird default for historical reason,